Università di Bologna
Dipartimento di Scienze Economiche Dottorato di Ricerca in Economia - XIX Ciclo
Esame …nale anno 2007
Foreign aid e¤ectiveness. An investigation of
the role of internal political constraints and
ownership
Dottorando: Lisa Montanari
Relatore: Prof. Paolo Manasse
1 Foreign aid e¤ectiveness. A survey of the literature. 1
1.1 Introduction . . . 1
1.2 Aid de…nition . . . 5
1.2.1 Kind of aid . . . 7
1.3 Aid and growth . . . 11
1.3.1 Some general issues . . . 11
1.3.2 Basic regressions . . . 13
1.3.3 Policy measures . . . 16
1.3.4 Main results . . . 19
1.4 The Principal-Agent Problem . . . 25
1.4.1 Country Ownership . . . 28
2 Internal political constraints 39 2.1 Introduction . . . 39
2.2 Basic model . . . 41
2.3 Internal political equilibrium with no foreign assistance 45 2.4 Internal political equilibrium in the presence of un-conditional aid . . . 49
2.5 Internal political equilibrium in the presence of con-ditional aid . . . 55
2.6 Conclusions . . . 59
2.7 Appendix 1: Aid repayment . . . 62 3
2.8 Appendix 2: Internal political equilibrium with aid
expropriability . . . 62
3 Political constraints under asymmetric information 67 3.1 Introduction . . . 67
3.2 The model . . . 68
3.3 Asymmetric information . . . 72
3.4 Con‡icting principals . . . 76
3.4.1 Complete information . . . 77
3.4.2 Asymmetric information over the government’s benev-olence . . . 79
3.5 Conclusion . . . 84
I examine the literature on foreign aid e¤ectiveness on recipient countries’ growth. Its contradictive results constitute the basis for the my investiga-tion. I present two main models which di¤er in the information set available to the agents that describe how the internal political constraints arising in the presence of a special interest group interact with the e¤ects of foreign assistance. The project is developed as follows. The opening of the thesis presents a detailed review of the Aid E¤ectiveness Literature, with a major focus on the third generation studies and the debate about the role the policy environment might play in increasing the e¤ects of aid on LDCs’economic growth. The shortcomings that commonly a¤ect the di¤erent strands of the empirical studies are highlighted and constitute the motivational base for fu-ture work to come. In particular I refer to the distinction between the e¤ects due to bilateral and multilateral aid (Ram, 2003; Rajan and Subramanian, 2005) and to kinds of aid programs having di¤erent timing impact (Clemens et al., 2004).
A lot of attention has been dedicated to analyse the impact of foreign assistance on the growth of recipient countries, but the evidence of the dif-ferent studies is in the end ambigous. Results can be classi…ed in three main strands. Some scholars argue that aid has been e¤ective in spurring growth on average, but with diminishing returns (Hansen and Tarp,2000;20011). Other
researchers assess on the contrary that it is not possible to identify a signif-icant relationship between aid and growth (Rajan and Subramanian, 2005). Finally, and more recently, a third view emerged. Aid has proved to have positive e¤ects on growth conditionally to some speci…c circumstances, such as a good policy or institutional environment (Burnside and Dollar, 2000; 2004), the non tropical geographic settlement of the country (Dalgaard et al., 2004) or the low vulnerability to shocks (Guillamont and Chauvet, 2002). I
1The amount of studies is huge. Here I only cite the most recent or in‡uencial ones.
do think that before adding results to the empirical side of the research a deeper theoretical investigation of the setting in which aid takes place is needed in order to derive more useful and clarifying insights.
That is why the core of the present work is theoretical and serves as a contribution to the diagnostic side of the aid research. I present a model in which the political authorities interact with a domestic special interest group that has an incentive to in‡uence policy decisions, following the path opened by Grossman and Helpman (1994). This setting is extended to the aid context considering the government of an aid recipient country that in-teracts also with a donor agency. The interest of the donor in the e¤ects of the policy implementation is direct or indirect, depending whether foreign assistance is conditional or not on the internal policy setting. Mayer and Mourmouras (2002) …rstly analyzed these circumstances in order to describe the advantages of conditional assistance strategies by multilateral agencies. Departing from them the aim of the present work is to identify the di¤erent equilibria (in terms of the combination of aid and distortion) that arise as the solution of this non cooperative game when the hypothesis about the cross marginal e¤ects of aid and policy distortion on welfare are allowed to change. When an increase in aid induces a lowering of the negative e¤ect of distortion on welfare this results less costly to bear for the government. The space for an incentive to accept aid while implementing a worse policy arises. The presence of internal political constraints such as the activity of a domestic lobby group may dampen the higher e¤ectiveness of foreign aid on the policy environment. When assistance is disbursed conditionally on the level of distortion chosen, the possibility of a deteriorated equilibrium to occur reduces, but in particular circumstances it might still be reached.
The results may contribute to the explanation of the ambiguity of empir-ical results regarding the sign and signi…cance of the aid-policy interaction term in growth regressions investigating the role of aid in enhancing LDCs’ economic performance over time. The aggregate measures of aid commonly used in these studies (O¢ cial Development Assistance or E¤ective Develop-ment Assistance) do not distinguish among kinds of assistance programs that potentially have di¤erent marginal e¤ects on the policy setting. Hence, the results might simply be biased by the matching of non omogeneous variables. It is interesting to note that the equilibrium combination of policy imple-mented and aid disbursed may turn to less distorted values if new strategies are considered. For instance, if the special interest group is somehow able to appropriate some of the foreign assistance entering the country (its utility
function positively depends on aid, along with the degree of distortion) the substitution e¤ect between aid and policy distortion is reduced and this leads to a more favourable equilibrium in terms of distortion. When aid repayment issues are taken into account it can be observed that, even if assistance bor-rowing is costly for the recipient country, some advantages may come from the reduction of the possibility of worse equilibria to arise.
Finally, the basic framework is extended to a context of asymmetric in-formation. The reforming attitude of the government is not observable by the lobby group before setting its contribution schedule aimed at in‡uencing the policy. The distortion produced in equilibrium is larger than the com-plete information one and positively depends on the size of the distribution over the “reform consensus parameter". The more volatile the government’s preferences are, the more the special interest group has to contribute in order to in‡uence political decisions, thus leading to a more distorted policy. The presence of an additional principal, the donor agency, is then considered. Its objective is opposite to the lobby one in terms of distortion. The donor’s assistance enters directly (and linearly) into the government’s welfare func-tion. In this framework aid is considered conditional and there is no direct e¤ect of foreign assistance on the public welfare. The burden of the con-tribution in equilibrium results biased towards the lobby. It represents the weak principal as its ideal distortion lies farther from the government’ one which is equal to the donor’s desired distortion. The International Financial Institution …nancing aid infact is considered as purely benevolent.
The aim of this project is to shed some more light over the proper identi…-cation of incentives underlying the foreign aid structure. The analysis of the interaction of all the actors involved: the donor agencies, the recipient coun-tries’governments and the domestic groups proctecting some vested interests by actively in‡uencing the government allows to draw some preliminary re-sults.
In presence of an active lobby group the fact that aid is particularly ef-fective in promoting welfare not only directly, but also trough its e¤ect on policies, may distort government’s incentives towards a higher level of distor-tion. In this circumstances, non trivially, if the lobby gains a special interest over foreign aid too, the distortion in equilibrium is reduced. The presence of asymmetric information regarding the preferences of the government towards reforms (what I de…ne the "consensus parameter", that can be thought of as well as the quality of institutions) makes distortions larger, the wider is the size of the distribution. The uncertainty over the government’s preferences is
identi…ed as a possible cause of the increase of the distortion in equilibrium and though as undermining the role of the conditionality of aid.
The parameter identifying the quality of institution here is exogenous, but the scope of further research is to endogenize it in order to design more accurately the interconnection between institutions and policies, collective action problem solution and …nally foreign aid e¤ectiveness.
Foreign aid e¤ectiveness. A
survey of the literature.
1.1
Introduction
Foreign assistance, in the sense we intend it nowadays, has been a post World World II phenomenon. The building up of major developed countries’multi-lateral institutions aimed at proctecting global economic stability (IMF) or at facing the problems of poverty in many countries of the world (World Bank) went step by step with the idea that monetary disbursements established a powerful instrument to overcome underdevelopment stigmata. O¢ cial strate-gic assistance ‡owing from industrial nations to colonized territories, com-mercial partners or military allies is an ancient costume. What distinguishes aid is an explicit development purpose that is by now a permanent feature of the modern global economy.
Since the end of the ’60s a large amount of attention has been dedicated to the analysis of the e¤ects of aid on LDCs’economic growth. A common classi…cation, …rstly introduced by Hansen and Tarp (2000), distinguishes the works in the literature into three generations’studies.
The earlier studies date back to the …rst ’70s. The core idea of the First Generation is that aid is considered only as an exogenous net increment to the capital stock of the recipient country. It is not treated as a component of the national income nor as an alternative source of increase in domestic con-sumption. The reason can be traced to the theoretical framework underlying the analysis, the Harrod-Domar (1946) growth model. Its assumptions rely
on a Leontief production function and on an excess supply of the labor in the economy. The only engine of an increase in output is the accumulation of capital which is the scarce factor of production. This implies production capacity to be proportional to the capital stock. Savings, both domestic and deriving from foreign aid, can a¤ect growth through investments that trans-late into accumulation of physical capital. The role of aid therefore consists of …lling the gap between required investment for growth and national savings. By the 60s, the Harrod-Domar model was already abonded as an instrument for growth analysis, but the development strategy implications deriving from the model were so handy that it survived as a development policy imple-mentation instrument for a long time. From an empirical point of view the tests to verify whether aid has a positive impact on growth concern the aid-saving relationship from which consequently derive the core aid-investment relationship. As Hansen and Tarp (2000) show in their survey of the First Generation regressions, the majority of works reported evidence of a posi-tive impact of aid on savings, but less proportional in size compared to the amount of aid received. In the 60s Chenery and Strout (1966) extended the Harrod-Domar model to a two-gap model, considering import capacity as an additional constraint on growth along with the saving constraint. Since foreign aid is supposed to equally lessen the …nancial and the trade balance gap simoultaneously, depending on which is the larger one (and binding one for the economy), the aid-saving interaction can be positive or not. When the trade balance gap is binding foreign assistance does not a¤ect growth via the savings channel. One of the two gap might bind in a country with respect to another or in one period time with respect to another inside the same country. This observation potentially justi…es the "undersize" of the saving to aid response.
The general opinion marking that period of research was indeed that aid was e¤ective in spurring growth. A few important drawbacks, anyway, were not taken into account. Firstly, the di¤erent empirical speci…cations of the aid-savings relationship carry with them distinct implications concerning the underlying saving behaviour in the economy and this is not generally recog-nized and explored. Secondly, and most importantly, no fungibility issues are ever considered. Aid was thought as pouring directly into productive invest-ment channels, but later evidence demonstrated that large amount of foreign assistance leads to an increase in the government’s consume expenditures1.
The …nancial gap approach was strongly critizised as being very naif in terms of proper incentives for development2. From a practical perspective o¢ cial
foreign assistance reaches the recipient countries by the hands of their govern-ments. Measuring the assistance need of a country as the di¤erence between its required investment and its domestic source of savings implicitly force a government willing to keep longlasting ‡ows active to divert aid use from saving to consumption in order to maintain the gap open and, accordingly, the foreign source of …nancing ‡owing in.
The Second Generation studies focuses both on the investigation of the direct relationship between aid and growth and on its indirect link via in-vestment. Papanek (1973) initiated a stream of models in which the di¤erent components of investment, including aid, are separated. Aid is no longer con-sidered as a fraction of total savings, but as an isolated component a¤ecting investments. Subsequent studies used reduced form equations on the same line to investigate the aid-growth relationship in cross-country studies.The results of these regressions are consistent both with the Harrod-Domar model and with the Solow growth model. In the latter framework imput substitu-tion is allowed while the capital-output ratio, the technology and the labour force rates are constant over time. The growth driving factor in the short run is still capital accumulation via investments.
Generally, these studies …nd a positive link between aid and investment, in line with the results of the First Generation. Moreover, a positive rela-tionship between aid and growth emerged conditional on savings having a positive e¤ect on growth. This allowed to con…rm a causality direction going from aid to savings, following up to investments and …nally to growth as standard growth models predict. Anyway, in these kind of analysis, the issue of potential endogeneity of the results is not still explicitely addressed.
For a more accurate accounting of endogeneity we have to wait for the Third Generation studies. This infact constitutes one of the speci…c char-acteristics of this latter series of empirical works. Panel data for a wide range of years and a great number of countries were used, allowing for better data reliability. The underlying theoretical framework relied on the endoge-nous growth theory. New variables measuring the quality of institutions and economic policies were included as controls in the regressions along with pre-viously used macroeconomic variables. In a series of studies, furthermore,
2The …nancing gap approach as a base for the determination of aid necessity was widely
aid was entered non-linearly in the growth regressions.
Since foreign assistance is typically distributed to poor countries in order to help them coming out from underdevelopment one could expect a negative link between aid and growth to be driven by the fact that lower growth countries tend to receive more foreign assistance. This issue was already noted in standard growth regressions for other variables measuring the state of the economy. In order to overcome this potential endogeneity problem most of the explanatory variables, including aid, are lagged one year. Some studies address the problem more accurately with an instrumental variable procedure in a two-stage least square.
The use of panel data, in all-but-one case divided in four-year periods3,
was completed to take account of country speci…c e¤ects and intercepts for each year were generally included to remove any world business cycle bias.
The standard controls related to macroeconomic environment are initial income, measured as the logarithm of real GDP per capita, which captures convergence e¤ects and aggregate money supply as a percentage of GDP, measured as M2/GDP and lagged one year M2/GDP, which identi…es the degree of …nancial depth. The measure of ethno-linguistic fractionalization proposed by Easterly and Levine (1997) and the number of assassinations per capita are also generally included to provide speci…c political environ-ment e¤ects. Di¤erent measures of the institutional and the policy set are considered in the di¤erent studies. I will come to their description later.
The intuition of non-linear e¤ects of aid on growth is addressed in two ways. The introduction of a squared aid term in growth regressions revealed very forceful. Its signi…cance with a negative sign revealed robust across all the studies. The presence of diminishing returns in aid appears as the most widely accepted result of all the aid e¤ectiveness literature, being robust to sample and controls modi…cations. The second strategy to identify non linearity is the inclusion of an interaction term between aid and the policy measure which captures the potential e¤ects of aid dependent on the policy setting.
These last generation studies can be split in two strands, according to the signi…cance of this interaction term. I will concentrate the analysis on the debate over the e¤ectiveness of aid tout-court or in the presence of a positive policy environment.
3The only exception is the work by Guillamont and Chauvet (2001) who divide the
In general the evidence about the aid-growth link is ambiguous. However, I can say that the majority of works tend to identify a positive and signi…cant relationship, whether direct or conditional on the policy set, between aid and growth on average.
1.2
Aid de…nition
The traditional measure of aid in the literature is the O¢ cial Development Assistance (ODA)4. It comprises o¢ cial …nancial ‡ows by developed country
governments and multilateral organizations. For a lending to be o¢ cially considered aid two features must hold. The motive of the …nancing has to be explicitely aimed at development purposes and funds have to be disbursed in the form of grants or concessional loans. Lending is considered concessional if, at full face value5, it has a grant element of at least 25% of the entire
disbursement. Even if the rate of interests applied to the lending are the market ones this gives rise to subsidized loans. Often the rates are lower than the predominant market interest rates. Indeed the loans bearing interest rates close to commercial ones are not included in the de…nition of foreign aid. The ODA measure includes both bilateral assistance, which is managed by agencies of donor governments and multilateral assistance which is funded by contributions from more developed countries and it is administered by international …nancial institutions, such as the World Bank, the International Monetary Fund or the United Nations Development Programme.
An alternative measure of aid introduced by Chang et al. (1999) has been largely emploied: E¤ective Development Assistance (EDA). It di¤ers from ODA mainly in two aspects. EDA only counts for the disbursements’grant component, but of all development …nance. In practice, the net present value of all the loans is considered regardless of their degree of concessionality. In this way EDA captures only the pure transfer of resources to recipient
4In 1961, the OECD set up the Developmnet Assistance Committee, which began to
publish annual statistics on aid.
5The face value is calculated, using a 10 percent discount rate. The calculation of the
grant element is done under the assumption that loan interest rates remain constant along the overall duration of the loan. This produces some shortcomings because prevailing market terms are ignored, The role of currency speci…c market rates is not taken into account as well as maturity di¤erentials.
countries. Additionaly, EDA’s adjusted grant equivalents are constructed with the use of market interest rates. The ODA measures the gross resource ‡ow, while the EDA considers its net ‡ow.
Futhermore, technical assistance, which is included in ODA, is excluded. Consultants and experts working in developing countries, in the majority of cases, are citizens of the donor countries. The reason underneath its exclusion in the new measure is that donors bene…t from payments received in return for the technical assistance supplied. On the one hand, the quid pro quo nature of this form of assistance makes its inclusion in the net assistance questionable. On the other hand, often the employment of foreign thechnical personnel is due to a lack of human capital in the recipient nations. From a theoretical point of view, the e¤ective amount of aid included in technical assistance should be valued by considering the cost the country would bear by hiring such skilled work on the free market. The presence of skilled foreign personnel may play a role in the formation of human capital and transfer of know how inside the countries. In this sense, the exclusion of technical assistance from aid measures seems more reasonable as far as it is tied to disbursements. Following this line of reasoning, anyway, would lead to the exclusion of all tied aid and not only technical assistance.
EDA certainly reveals a more accurate measure when we come to make comparisons of aid ‡ows across donors and recipients, since ODA tend to overestimate the aid ‡ows to countries receiving mostly loans6. On the other
hand, aid ‡ows of mostly grants giving donors tend to be underestimated7. When the focus of the analysis turns to empirical studies addressing the e¤ect of aid on growth using panel data however, since ODA and EDA are highly correlated through time , the di¤erence in the results by the switch from one measure to the other is not so relevant. Whatever the measurement used, the aid value is then generally converted into constant 1985 dollars8 and divided
into real GDP from Penn World Table 5.59.
Renard and Cassimon (2003) propose a few alternative approaches to aid
6The grant elements of concessional loans are overstated especially if they are
denom-inated in the recipient country’s currency, because the related interest rates are typically lower than 10 percent considered in the ODA measure.
7This occurs especially for donors’currency denominated loans.
8The transformation is done using the IMF’s world Export Unit Value Index. 9From Summers and Heston (1991).
measures. They distinguish the donor’s perspective from the recipient’s per-spective. For the donor, development assistance involves some costs. Three ways of calculating the costs are considered. Firstly, gross budgetary costs which derive from the face value of the loans and include gross salary paid to technical assistance. Secondly, net budgetary costs in which net transfers on loans and net salaries are considered. Basically, reverse …nancial ‡ows coming from aid, such as taxes on the salaries of the donor country citizens working as technical assistants and debt service paid by recipients on previ-ous loans are excluded. Finally, what they de…ne "economic cost" takes into account the opportunity cost of not using the resources given to development aid in the donor economy.It is the most complete expression, but it is hardly impossible to measure since it requires a comparison between the returns on alternative investments and the bene…ts donors gain from development assistance. These range from the consolidation of trade partnerships to the management of military threats, from the control of world political stability to the containment of illegal migration and so on. Constructing a measure of these bene…ts would probably lead to more drawbacks than using the avail-able aid measurements. From the recipient’s perspective aid is evaluated in two ways. The acquisition value of aid represents the cost of obtaining the same goods and services provided through foreign assistance on the world market. The …nal use value of aid aims at calculating the economic net present value of projects funded with concessional loans. Evaluating each aid component on a continous basis anyway requires huge e¤orts and it con-sists of a non speculative procedure only on an ex post basis. Renard and Cassimon’s proposals are interesting from a theoretical point of view. They specify "what we should be measuring, and why we don’t", but they are di¢ cult to be translated into meaningful data.
1.2.1
Kind of aid
Aid appears as a very generic entity once we go into depth and understand how many di¤erent components it encounters. The main distinction is be-tween “project aid” and “program aid”.
“Project aid” involves activities over which donors try to exercise a de-tailed and ongoing control. Aid projects generaly refer to funding for speci…c investments, for instance a road or capacity building for environmental man-agement. On average they last about three years.
“Program aid” instead comprehends budget support, adjustment loans, debt relief, and sector-wide action programs. Program aid refers to funding given to a government to fund general public expenditures. It is usually provided on the basis of an agreement on how money will be spent or of speci…c reforms.
A third voice, however, deserves mention: "emergency aid", including food and prime-care commodities.
Distinguishing emergency from non-emegency disbursements is easy since Development Assistance Committee (DAC) statistics provides this informa-tion. To distinguish non-emergency aid into project and program ‡ows is less straightforward. From a measurement point of view, project aid is sector-speci…c while program aid is not allocable to any particular sector.
Di¤erent kinds of aid might a¤ect growth in di¤erent ways. The only empirical study that attempt at isolating the e¤ect of di¤erent kinds of aid is Clemens et al.(2004)10. The underlying idea is that di¤erent kind of aid may have e¤ects on growth at di¤erent distance time. They separate aid into three categories: emergency and humanitarian aid (very short-term), aid that a¤ects growth only after a long period of time (long-term) and aid that is directly aimed at a¤ecting growth within a period of four years (short-term). Very short-term aid, like food aid, is likely to be negatively associated with growth, since aid tends to increase fast at the same time growth falls following an economic shock. Long-term aid, such as aid for health, education, the environment, and democracy support may be di¢ cult to identify in a short-run horizon. Finally, short-term aid, such as aid for transports and infrastracture or agriculture should be the most e¤ective in spurring growth. A strong positive relationship between the third type of aid (which is nearly half of the overall aid) and growth emerged and proved very robust to a variety of checks. As expected, the relationship with the other types was less signi…cative. In my opinion, this seems the right direction to
10There are some previous studies that attempt at disaggregating aid ‡ows, but they
consider assistance to a speci…c country or divide aid in wider categories. Owen and Hoddinott (1999) distinguishe the e¤ects of development aid and humanitarian aid on household welfare in Zimbabwe. Mavrotas (2002) …nds that program aid, project aid and technical assistance all had a negative e¤ect on growth in India,in the period from 1970 to 1992, while in Uganda only technical assistance had a negative impact. Gomanee et al.(2002), analyzing the e¤ect of aid on growth in Africa, substract from ODA food and technical assistance as they have an impact on di¤erent time-horizons. Pettersson (2004) tries to divide aid among sectors, depending on whether aid is perceived to be fungible in the specifc sectors and …nds no di¤erence in the growth e¤ects.
follow to obtain more sensitive and uniform results about the e¤ectiveness of aid.
Another distinction in the overall aid has to be made. ODA includes both multilateral and bilateral aid. Multilateral assistance is provided by international multilateral agencies and bilateral aid is directly …nanced by the donor countries through government agencies. International Financial Organizations are also …nanced by developed countries, but the motivations of this indirect assistance and the o¢ cial bilateral one may be very di¤erent. Political, commercial and military interests are often at the basis of bilateral funding. In such cases the promotion of recipient countries’growth is not the core objective Some bilateral aid is tied. It is disbursed under the condition of being used to purchase goods and services from the donor country. Aid tying has been object of many critics. It seems to reduce the value of assistance by about 25 percent. Because of that there has been an evident decrease in tied aid through time in OECD countries. After a consistent amount of foreign aid in the 70s and in the 80s, the following decade was characterized by a sharp decrease in aid ‡ows. The trend of aid is shown in Figure 1 for ODA/GNI from the 60s. Most of the decline is due to bilateral aid. Fiscal problems in OECD countries led governments to contract expenditures. Even though foreign assistance did not represent a major component of government spending, it was often curtailed. The end of the Cold War made the necessity of a Western driven development less stringent. As a consequence, foreign assistance to "buy" political allies suddenly fell. Additionally, private capital ‡ows directed towards developing countries sharply increased in the early 90s. Foreign investment partly compensated the needs of LDCs, but it was concentrated in some particular countries or regions and, after the asian …nancial crisis in 1997 it started diminishing. The turndown in aid has been reversed at the beginning of this decade. On 2nd March 2005 Paris Declaration on Aid E¤ectiveness was held with the speci…c purpose of pushing the management of aid towards a more e¤ective path in the achievement of the Millenium Development Goals. These goals aim at cutting in half the proportion of people living in absolute poverty within year 2015. Many donor countries renewed their commitment to reach an amount of resources dedicated to development assistance of 0.7% of their GNI. However, the rise in aid ‡ows accelerated after the 9/11 attack against the United States. New resources dedicated to development assistance in recent years have as a background the "war on terrorism".
F igure 1: Long-term trend in DAC ODA-GNI ratio (1960-2004) Source: OECD DAC ( DAC projections for the period 2006-2010)
(1.1) While altruism can be seen as part of the intent of foreign assistance for both multilateral and bilateral aid, the latter kind of aid is more related to the donors’ strategic interests11. These interests may also involve
posi-tive aspects. Whatever the reasons underneath long-standing relationships between countries (colonial past, commercial partnership, military alliance), they often bring to some similarities in (business) language, technologies (and related labour skills) and the legal and institutional setting which may lead to higher returns of aid ‡ows. The e¤ect of the di¤erent components of aid on growth may reveal some di¤erences. Bilateral aid is typically larger than multilateral aid. At least, it should be subject to di¤erent controls in the analysis or, as only Ram (2003) does, it should be included in the analysis with a di¤erent weight with respect to multilateral aid. In the following sec-tion I describe in details the way aid-growth regressions are speci…ed. Here I anticipate that when the total aid variable is included in the growth regression
implicitly equal weights are attached to its components (bilateral and multi-lateral aid). Ram shows that if the two components are left unconstrained (by entering them separatly in the regression equation12) their parameters take
values of opposite sign, suggesting that bilateral and multilateral aid play an opposite role on growth. I postpone the discussion over the speci…c e¤ects to the next section, but I wish to emphasize that further analysis would de-serve mention on this issue. Clemens et al.(2004) and Ram (2003) show how disaggregating aid into distinct components, di¤erent time-impact aid and multilateral and bilateral elements, respectively, produce meaningful results in terms of signi…cance of the aid variable coe¢ cient. One recent study by Rajan and Subramanian (2005), checks for both kinds of disaggregations and does not …nd any signi…cant additional information with respect to the use of total aid. Ambiguouity comes around again. I do think, anyway, that their line of research is the one to follow in order to obtain a deeper insight into the aid e¤ectiveness topic. The …rst natural step, which will be the scope of further research, is to jointly address the two di¤erent decompositions of aid, by analyzing the e¤ect of di¤erent time-horizon multilateral aid on growth and repeating the exercise for bilateral aid.
1.3
Aid and growth
1.3.1
Some general issues
The reason why the relationship between aid and growth has attracted so much attention is due to the recognized e¤ect of economic growth in reducing poverty and improving social indicators. When income of the poor popula-tion rises, the opportunity to improve health, educapopula-tion and living standards follows. All foreign assistance is ultimately aimed at achieving these objec-tives.The e¤ects that aid has produced in time and across countries anyway is far from being omogeneous.
As we brie‡y mentioned in the introduction, in the earlier times of devel-opment assistance, the lack of savings and foreign exchange for investment were considered the the key responsible factors holding back poor countries. The consequent rationale for foreign aid strategy was to lend funds to
coun-12The possibility of studying the e¤ects of the two components separately was provided
tries in order to overcome a “saving gap”to …nance necessary investment and a “trade gap” so that imported machinery could be the engine of that in-vestment. Development agencies worked with a “two gap" model that made imports and investment in physical capital the driving force of growth. The " gap" strategy was largely implemented in the Soviet countries through development planning, but the collapse of the system showed that invest-ment alone does not assure growth. In the 1990s emphasis has shifted from investment to other objectives, such as the improvement of institutions and policies which promote growth, human capital development and technological progress. These strategies were in line with the …ndings of the new growth theory that attempts to identify the underlying factors of growth, besides the physical capital accumulation. The later strand of the growth litera-ture, moreover, turned the attention to the institutional and policy setting of a country as a major determinant of its economic performance through time. The importance of a stable macroeconomic environment comes to be considered essential for economic growth to take place. High in‡ation is bad for investment and growth (Fischer 1993). In the same way, large …scal de…cits delay growth (Easterly and Rebelo 1993). The openness of markets is crucial. Most trade liberalizations accelerate growth (Sachs and Warner 1995). Fiscal, monetary, and commercial policies determine the macroeco-nomic management level of a country and there is wide evidence that a good macroeconomic management provides a growth-prone environment. Good institutions and a solid economic management are important also at the mi-croeconomic level. The strength of private property rights and of the rule of law and the quality of the civil service a¤ect long-term growth (Knack and Keefer 1995). Corruption in the public bureaucracy negatively a¤ects growth (Mauro 1995).
Anyway, observing the simple relationship between aid and per capita growth can be misleading. Some developing countries which received large amounts of aid have grown slowly ( Zambia, for instance) while others have grown rapidly (Ghana). Certainly, even when aid has a positive e¤ect on growth it is not the only factor driving growth. Other variables that have an impact on growth must be taken into account. Furthermore, countries which experienced a poor growth are very likely to be among the ones which received assistance. This raises the doubt that the causal relation might run from low growth to an increase in aid, rather than in the opposite sense. It would not be correct to interpret a negative relation, for example, as evidence that aid reduces growth. The endogeneity problem has to explicitly addressed.
1.3.2
Basic regressions
Before describing the main and con‡icting results obtained in the Aid E¤ec-tiveness Literature I describe a representative regression of the kind used in almost all the cross-country empirical studies.
The dependent variable is the annual growth rate of per capita GDP (g), averaged over the benchmark period. The sample is a panel of countries which covers an average period of 25 years13. Typically data are divided in
four-year periods. The growth rate is a function of exogenous conditions (X), the level of policy (P), the level of aid relative to GDP (A), the level of aid squared , and the interaction of the policy variable and aid :
g = 0+ 1X + 2P + 3A + 4A2 + 5A P + 6d + u (1.2) The coe¢ cient on the interaction term, 5, addresses the hypothesis that the e¤ectiveness of aid depends on the policy environment, while the coe¢ -cient on the quadratic term, 4 picks up any diminishing returns to aid. The
coe¢ cient on aid, 5, may be positive, negative or zero depending upon the way policy in‡uences aid e¤ectiveness. The coe¢ cient 3 indicates the direct e¤ect of aid on growth. The constant 0 is usually entered as a measure
of initial GDP per capita in order to capture any country …xed e¤ect. The coe¢ cient 1 gives us information about the role played by the various con-trols in determining growth. Concon-trols vary across the regressions or address the same variable with di¤erent measures. I below cite the most commonly used. In order to address the problem of endogeneity, the controls are often lagged one period. Further on, regional dummies are entered. Typically, Sub-Saharan Africa and East Asia14 whose growth rates might be a¤ected
by common determinants of the region that , if no isolated, could bias the coe¢ cients of the regression correlated with these same regional features. Fi-nally, the error term (u) is entered. The standard controls most commonly used are
- Private investments - Government investments
13In previous studies covers the period from 1970 to 1993. In later works the sample
was expanded.
- Domestic investments - Government expenditures - Private net in‡ows
- Human capital
- Population growth rate - Terms of trade
- Real exchange rate
- Ethno-linguistic fractionalization - Assassinations per capita
- Assassinations Ethnic fractionalization - Institutional quality
- Revolution - Financial depth
- Fraction of land in the tropics
In spite of explicitly addressing endogeneity, Boone (1994) …nds no re-lationship between aid and growth, even after controlling for a range of in-stitutional and political factors. This can be considered the initial paper of the Third Generation of studies about aid e¤ectiveness. It comprises all the typical features of the latter stream of analysis. In addition, its discouraging results generated a renewal of interest for the topic. Boone’s stronger result regards the positive e¤ect of aid on government’s consumption to the expense of growth-enhancing investments. He develops these results for the 1971-1990 period, suggesting that missing development was due to the fact that poverty is not the direct consequence of the lack of stock capital. Further on, until the beginning of the 90s aid giving was strictly related to poverty conditions of the recipient countries and completely unrelated to changing in internal policies. The resulting distorsive e¤ect on incentives to implement reforms for the national authorities is nowadays well recognized. Boone’s analysis
is consistent with a model setting of di¤erent political regimes searching to maximize the utility of their supporting group, whether this is an elitarian group, an egalitarian one or a "lassaiz-faire group" oriented to policies aimed at reducing distortionary taxes. His …ndings are in line with the framework of an elitarian regime exploiting fungible aid to maintain political control. All the systems support the elites at the margin, but, other things equal, considering the basic development indicators, more libertarian governments generate a 30% lower infant mortality rate with respect to other kinds of governments. Boone emphasizes that, although aid does not prove e¤ective in enhancing growth, it may bring other positive e¤ects on development.
The picture changes thanks to a very in‡uencial paper by Burnside and Dollar (2000)15 in which the sample is divided between good management
and poor-management countries. The proposition that aid has no e¤ect in the presence of weak incentives is confermed by the empirical …ndings. For countries with poor management, whatever the amount of aid, growth was null, or negative. For countries with good management growth was posi-tive, with size di¤erences depending on the associated amount of aid. The group good-management and low-aid countries had an average growth of 2.2 percent per capita, while the group of good-management and high-aid group countries grew more rapidly at 3.7. Excluding middle-income coun-tries, which receive little aid, from the sample makes the e¤ect of aid even stronger. Considering the growth rates in the 90s to be coherent to Burnside and Dollar’s sample period, countries with "good policies" that have received a lot of aid and experienced a good performance include Bolivia, Honduras, Ghana and Mali. Case studies provide some speci…c insights. A comparative one targeting aid e¤ectiveness in Bolivia, Costa Rica, and Nicaragua com-missioned by the Colombia’s Ministery of Planning supported Burnside and Dollar’s conclusions. In Nicaragua, a large …scal de…cit and high in‡ation rates accompanied a very poor e¤ect of aid on economic growth. In the op-posite way, aid was largely e¤ective in Bolivia, which successfully completed a reform program in the 1980s. The following quotation by Minister Lopez (1997) about the three case studies re‡ects a similar view regarding the role of policies :
Foreign aid in itself is neutral with respect to development, for its positive or negative e¤ects depend on government policies.
Ef-15The paper was published in 2000, but since 1997 the working paper widely circulated.
fects on economic development will tend to be positive when aid is used to build up capital or to …nance public investment that con-tributes to the pro…tability of private capital, or for human capital development. Fiscal policy should generate government’s current savings, so that both domestic and foreign resources …nance pub-lic investment. If this does not happen, foreign resources may end up …nancing the government’s current expenses and not in-vestment projects, as happened in Nicaragua. The relationship between aid’s positive e¤ects and good domestic policies always holds, even during adjustment (Lopez 1997).
The work by Burnside and Dollar (2000) received severe critiques, but its key role in the debate over the way aid can turn to be e¤ective for growth deserves a deeper understanding. Firstly, we need to de…ne which measure of policy was considered.
1.3.3
Policy measures
What are good policies? For their study of aid and growth, Burnside and Dollar created a policy index based on the Sachs-Warner measure of openness, the in‡ation rate and the government budget surplus. By running a …rst regression of the three variables over per capita income growth they obtained coe¢ cients that were used as the weights of the index. Their data set cover 56 aid-receiving developing countries. Growth, aid, and the other variables are averaged over four-year periods (starting with 1970–73 and ending with 1990–93).
The choice of this policy index was strongly criticized. In general it seemed that a sensitive measure of a good policy environment could hardly be obtained by such a limited measure. More in the details, it was observed that these macro focused variables are likely to be endogenous with respect to growth.
Further studies adopt a di¤erent broad measure for the quality of pol-icy that partially takes into account the institutional setting of a country: the CPIA. The Country Policy and Institutional Assessment estimates the quality of a country’s present policy and institutional framework. "Quality” refers to the extent to which a policy contributes to promote poverty reduc-tion, sustainable growth and the e¤ective use of development assistance. The
rating is built on a 1 to 6 scale. Since 2005 the survey are available to the public, but as before the exact data were not published it was necessary to use a proxy measure for the CPIA. The components of the index are divided in four main areas, as follows:
A. Economic Management: 1. Macroeconomic Management 2. Fiscal Policy 3. Debt Policy B. Structural Policies 4. Trade 5. Financial Sector
6. Business Regulatory Environment C. Policies for Social Inclusion/Equity
7. Gender Equality
8. Equity of Public Resource Use 9. Building Human Resources 10. Social Protection and Labor 11. Policies and Institutions for
Environmental Sustainability D. Public Sector Management and Institutions
12. Property Rights and Rule-based Governance
13. Quality of Budgetary and Financial Management
14. E¢ ciency of Revenue Mobilization 15. Quality of Public Administration 16.Transparency, Accountability and
Collier and Dollar (2002) repeated Burnside and Dollar’s exercise using this di¤erent measure for the quality of policy (CPIA ,1997) and considering an alternative data set. The idea was to concentrate on the 90s to avoid results driven from the switch in aid amounts due to the end of the Cold War. This allowed anyway to include many more countries in the sample. The main objective of the paper is to identify the di¤erence between an ef-…cient and the current aid allocation, but the …ndings prove supportive of the signi…cance of the aid policy interaction term. Among the controls to capture initial conditions along with initial income also a measure of insti-tutional quality is considered: the ICRGE from Knack and Keefer (1995). It is an index based on the evaluations of …ve di¤erent institutional indica-tors made by the private international investment risk service, International Country Risk Guide. In particular they are the quality of the bureaucracy, corruption in government, rule of law, expropriation risk and repudiation of contracts by the government. The institutional quality variable is included as time constant. Hence, a strong assumption of constancy and exogene-ity of institutions is needed The authors compute a regression adding an aid institutions term in order to control whether the aid policy term might be proxying for this variable. The aid-policy interaction term continues to be statistically signi…cant and positive, suggesting a distinct role of policies and institutions in a¤ecting returns to aid.
A di¤erent perspective is followed by Burnside and Dollar (2004). They further expand the policy measure to embody the e¤ects of institutions. After evaluating the di¤erent measures used in the empirical growth literature to determine the quality of institutions16 they choose an index of institutional quality which is computed by standardizing and averaging all the di¤erent institutional variables used in the literature in the second half of the 90s17.
The index is showed to combine the information of the ICRG rule of law index and the Freedom House democracy index18. By using an instrumental
variable tecnique in order to address endogeneity problems coming from the
16See Knack and Keefer (1995), Hall and Jones (1999), Rodrik (1999), Acemoglu et
al.(2001), Rodrik et al. (2002) and Dollar and Kray (2003).
17See Kaufmann, Kraay and Zoido-Lobotan (1999) and also Easterly and Levine (2003). 18The ICRG index ranges from 1 to 6, the higher number indicating better
institu-tions. As I mentioned before, the rule of law is one of its …ve components. The Freedom House democracy index ranges from 1 to 3, the lower number indicating more democratic institutions.
subjective nature of the institutional quality variables19, Burnside and Dollar
conclude that " with a new data-set covering a di¤erent period (the 1990s) and using a di¤erent index of institutions and policies, we …nd the same basic pattern that we identi…ed in Burnside and Dollar (2000): the combination of good instituions/policies and external aid appears to lead to more rapid growth".
The latter two mentioned studies somehow contradict each other. The benchmark period of the sample is the same, but the broader measure of policy suggests a key role of institutions in aid e¤ectiveness that was absent in the previous study (the interaction term between institutional quality and aid is only marginally signi…cant and its economic e¤ect is small and negative). Results seem sensitive to the choice of the policy index used. This is one of the main critiques targeted to the "conditional strand" of the aid e¤ectiveness literature. A wider view over the main results needs to be considered.
1.3.4
Main results
Economic and development objectives fostered by foreign aid are very di¤er-ent. Building infrastructure, supporting sectors that prove locally produc-tive, such as agriculture, circulating ideas and new technologies, improving education, health and political systems, providing subsistence consumption, especially food humanitarian crises, helping stabilize the economic systems striken by economic shocks and so on. Despite these broader objectives for aid, the key idea has always been that more aid is expected to lead to faster growth. At a general level, there is no apparent simple relationship between aid and growth. Some analysts argue that once endogeneity factors are taken into consideration, a positive relationship emerges. Others conclude that aid works well under certain circumstances, but has no e¤ects at all in others. Such circumstances might be related to the policy and institutional environ-ment, to geographical features or the economical and political setting. In this view, while the average trend is important, the variance around the trend and the determinants for that variance are also critical in understanding the core underlying relationships.
A lively debate animated the literature and the development strategy circles for a long time. There is general agreement on some broad issues. Even aid "non believers" agree that aid has been successful in some countries (Botswana and Indonesia, for instance). It has helped to improve health by supplying essential medical care and providing emergency relief after natural disasters. Similarly, "aid believers" agree that a gross amount of aid has been wasted, for example in highly corrupt political settings, such as the Marcos regime in the Philippines. It can have adverse incentives on economic activity. The conditions under which aid works or does not work are still a hot topic in the development debate. Empirical evidence brings ambiguous results. Di¤erent studies reached di¤erent …ndings, depending on the time frame, the countries involved, and the theoretical assumptions underlying the research. Results can be divided into three main strands.
The optimistic view supports the e¤ectiveness of aid, despite recogniz-ing its limits. Aid has a positive relationship with growth on average across countries, but with diminishing returns as the amount of aid increases. The most commonly reason ascribed to explain the presence of diminishing re-turns to aid is a limited absorptive capacity by the recipient country. The lack of adequate human capital and weak infrastructures lead to the waste or suboptimal utilization of a part of assistance resources when the amount disbursed increases.
The underlying idea, supported by the First generation studies, is that aid augments saving and spurs investment. It sums up to the capital stock. Poor countries would be unable to generate su¢ cient amounts of saving to …nance the investment required to prompt growth. The poorest countries may even be stuck in a poverty trap in which their income is too low to generate the saving necessary to initiate the process of sustained growth. This is the message o¢ cially sustained by Je¤rey Sachs as Un advisor20. Aid
is argued not to prove e¤ective in spurring growth essentially because the quantity of assistance has not been su¢ cient to induce the necessary jump out of the poverty trap. Large amounts of aid given through time may reveal less e¤ective than big volumes of aid properly targeted at right times. There are di¤erent channels able to connect aid to a positive economic performance. The endogenous growth theory gives the analytical foundation for an increase in worker productivity driven by investments in health or education. Aid
20See Sachs et al.(2004) in the document "Ending Africa’s Poverty Trap" prepared in
could provide a transfer of technology or knowledge from rich countries to poor countries through technical assistance, or through direct transfer of technologies, for instance the introduction of fertilizers in the agricultural sector. These studies do not conclude that aid has always worked in every country, but that on average and controlling for other factors, higher aid leads to larger growth.
The opposite view re‡ects results showing that aid has no signi…cant rela-tionship on growth and in case it were signi…cative it may actually undermine growth. This position can be traced back to Peter Bauer who was perhaps the most in‡uencial supporter of this idea of development21, although he
never provided systematic empirical evidence to support his argument. Sev-eral empirical studies draw the conclusion of no relationship between aid and growth. Researchers have suggested di¤erent reasons to interpret their …ndings. Aid can distort incentives for private sector activities. It can spur in‡ation and cause a real appreciation of the domestic currency leading to “Dutch disease” e¤ects22. Hence, to the extent to which tradable activities
are a key source of productivity gains, long-term growth may su¤er. Aid ‡ows can prompt the augment of factors implied in the services supporting aid projects, drawing workers and investment away from other productive activities. Food aid, in particular, can sometimes induce a fall in prices in the internal food production market. Furthermore, aid simply could be wasted, enlarging government’s consumption or it could augment corruption of the bureaucratic apparatus. Aid might help keeping bad governments in power, thus delaying reform implementation. Some argue that aid provided to countries in con‡ict times indirectly …nances the continuation of the con-‡ict, generating more instability. Hansen and Tarp, in a detailed survey of the Aid E¤ectiveness Literature, show how the number of published empiri-cal studies that have found no relationship between aid and growth are just a few compared to the total number of studies. Most of them use restrictive models, imposing a linear relationship between aid and growth and ruling out by assumption the possibility of diminishing returns. Most also only examine aggregate aid, implicitly assuming that all aid has a similar impact on growth.
As I mentioned before, in the middle of the1990s researchers began to
21See Bauer (1972).
22The appreciation of the exchange rate reduces the pro…tability of the production of
check whether aid might support growth with diminishing returns. A large group of studies that allow for diminishing returns have found a positive rela-tionship. This is probably the more robust result of all the Aid E¤ectiveness Literature. The non-linearity of the aid-growth relationship anyway were addressed also in an alternative way, by testing for conditional relationships. This is the last and most recent strand of research about the topic. After con-trolling for speci…c variables and allowing for diminishing returns, a positive relationship between aid and growth emerges, albeit with important variance around the trend line. E¤ectiveness of aid may be conditional to features of the recipient countries. This line of research was anticipated by Isham et al.(1995) whose results showed that World Bank projects had higher rates of returns in countries with stronger civil liberties. Burnside and Dollar (2000) concluded that aid stimulates growth in countries with good policies. As we discussed in the previous section, this work had a major role in prompting many scholars to investigate aid e¤ectiveness with this new approach.
This “conditional” strand of the literature has taken into account di¤er-ent characteristics of the recipidi¤er-ent country that might a¤ect the aid-growth relationship. In Table 1 the main results are summarized. The interaction between aid and some policy variable is the most tested. I described more in detail the di¤erent policy measures in the previous subsection. Collier and Dehn (2001) analyze the e¤ects of large negative terms of trade shocks. They …nd that while such shocks strongly damage growth, additional aid reduces the size of growth losses. Chauvet and Guillamont (2002) consider a wider measure of shocks. They …nd a positive coe¢ cient of the interactive term between aid and the economic vulnerability to exogenous shocks (de…ned as environment). In countries striken by climatic or external shocks aid reduces the growth decline and allows for policy stability to be maintained. The "en-vironment" index includes four components. The instability of agricultural production is the proxy for climatic shocks, the smallness of the population size for the exposure to shocks. The other two measures are the instability of the exports earnings and the trend of the terms of trade. The index is a weighted average of the four variables. Collier and Hoe- er (2002) analyze a sample of 17 countries which experienced a civil war during the …rst decade of their post-con‡ict economic recovery. Their results lead to the observa-tion that in the second part of the decade the absorptive capacity for foreign assistance is systematically higher. This does not occur instead in the very short-run (the …rst three years after the con‡ict). Finally, Dalgaard et al. (2004) suggest that during their 30 year sample period aid resulted to be
less e¤ective in the tropics. They do not o¤er any meaningful explanation of the phenomenon. The negative relationship between the tropical area and growth is a robust result in the literature (Easterly and Levine, 2003), but also adding the aid tropical area fraction in the aid-growth regression draws to a much more robust coe¢ cient than considering the aid policy interaction term.
Study Period Signi…cant interaction term Burnside&Dollar(2000) 1970-93 aid B-D policy index Collier&Dehn(2001) 1974-93 aid B-D policy index aid negative shock Guillamont&Chauvet(2002) 1970-93 aid environment Collier&Dollar(2002) 1974-97 aid CPIA
Collier&Hoe- er(2002) 1974-97 aid CPIA post-con‡ict Dalgaard et al.(2004) 1970-97 aid tropical area fraction Burnside&Dollar(2004) 1990-99 aid institutional quality
T able 1:"Conditional" strand studies
These "conditional" studies, if relieved on, clearly have important devel-opment policy implications. The idea that aid works better in countries with good policies and institutions spread among donors, because this line of re-search re‡ects the beliefs of the majority of development operators working on the …eld. This approach can also explain the volatility of the aid e¤ects across countries.
Unfortunately, the …ndings of this strand of the literature have proved very fragile. Their procedure relies on an interaction term between aid and the variable considered, but many of the interaction terms are not robust to changes in the speci…cation. Easterly et al. (2004) …nd that the original Burnside and Dollar results are highly data dependent. They are driven by the presence of …ve outliers and they do not hold up to en enlargement of the sample period and other robustness checks. Ram (2004) disaggregated the aid variable into a bilateral and a multilateral component. Both kinds of as-sistance revealed signi…cant coe¢ cients, although of opposite sign, but their interaction term with the policy variable did not. These …ndings suggest that, on average, bilateral aid has been e¤ective in promoting growth, while multilateral aid has had a negative e¤ective. This view is opposite to what it is commonly perceived. Attempting to give an explanation for the results, the author argues that the special link usually present between the donors
and the recipients might have positive spill overs over aid returns. In general, donors have quite a deep knowledge of the recipients’economies due to his-torical ties. This yields to the adquirance of experience and skills speci…c to the recipient countries which often come along with linguistic a¢ nities and similarities of the institutional structures. Rajan and Subramanian (2005) repeat the exercise and do not reach the same conclusions. The works ad-dressing the topic in the literature anyway are just a few. It would deserve further investigation and testing.
Since the number of works in the AEL has been huge23, some economists
attempted to build a survey of the literature with the intent of comparing the di¤erent studies by deriving omogeneous measures in terms of samples and controls speci…cations.
The more complete ones are Hansen and Tarp(2001), Roodman(2004), and recently Rajan and Subramanian (2005). Roodman (2004) …nds most of the "conditional" studies to be relatively fragile. Sample variations, di¤er-ent speci…cation of the "conditional" variable and the inclusion of di¤erdi¤er-ent controls in the regressions tend to generate alternative results. Only the con-clusions of Dalgaard and Tarp (2004) over the aid-tropic link seem robust. Hansen and Tarp and Rajan Subramanian reach the same conclusion, but with opposite e¤ects. The earlier study supports the idea that aid is e¤ective in any country, despite the quality of its policies, while the more recent study sheds a new shadow over aid e¤ectiveness. No evidence of a direct system-atic relationship between aid and growth is found. Furthermore, no indirect link through the geographical or policy environment is veri…ed and even the di¤erences in the forms of aid (bilateral vs multilateral or short-impact vs long-impact) seem to play any role. We are back to the original question: "Does aid really work? If not, why?"
To sum up the aid and growth research, it appears that aid has been successful in some countries but not in others. The overall trend is a subject of debate, the results in the literature are not omogeneous. ( although most researchers, as a number, have found a positive relationship). The analy-sis turned to the investigation of the conditions under which aid has the largest impact on growth and of what types of aid are most e¤ective. The only commonly recognized robust …nding up to now seems to be that aid
23Just to mention a few more: Hadjimichael et al.(1995); Durbarry et al.(1998);
Dal-gaard and Hansen (2000); Hansen and Tarp (2000); Lensink and White (2001) and Clemens et. al.(2004).
has diminishing returns (the coe¢ cient of the aid2term in the regressions is
negative).
I do believe that new hints coming from the growth literature could spill over to the AEL. For instance, Rodrik (2006), commenting on the World Bank’s document "Economic Growth in the 1990s: Learning from a Decade of Reform" proposes a new strategy for targeting the limits of development: " Reform e¤orts need to be selective and focus on the binding constraints on economic growth rather than take a laundry-list approach à la Washington Consensus...In a low-income economy, economic activity must be constrained by at least one of the following two factors: either the cost of …nance must be too high, or the private return to investment must be low. If the problem is with low private returns, that in turn it must be due either to low eco-nomic (social) returns, or to a large gap between social and private returns (low private appropriability)". The choice of the policy to be promoted in a country should be targeted to the most binding constraint for that economy, otherwise the positive e¤ect of the policy itself in enhancing growth could be undermined. When investment is constrained by poor property rights, for example, improving …nancial intermediation will be of little help. When it is constrained by the high cost of capital, improving institutional qual-ity will not be useful. Hence, the …rst step for a well-targeted development strategy consists of undertaking a diagnostic analysis to identify the most binding constraints on economic growth in a given setting. If policies prove ine¤ective only because they are directed to the wrong sectors of the econ-omy, consequently even aid given to promote such policies will turn out as ine¤ective in spurring growth. In di¤erent settings anyway, "…nancing" the same kind of policies could be very helpful. With my analysis, I intend to bring a further contribution to the diagnostic side of the aid literature, but before another feature must be taken into account in order to understand all the potential limits to proper aid e¤ectiveness.
1.4
The Principal-Agent Problem
As I said, multilateral aid is perceived to be more e¤ective than bilateral aid. The same thing occurs for untied aid with respect to tied aid. How-ever, the problems can be analyzed under a more general framework with respect to the programs characteristics and the degree of internal coherence
between their objectives and the instruments e¤ectively implied to realize them. The donors that have large bureaucracies, that do not coordinate with other donors, or that have poor monitoring and evaluation systems are thought to undermine the e¤ectiveness of their aid programs. Recently, there has been a renewed interest for an increasing “country ownership”and a broader participation of the government and the community groups in recipient countries in choosing priorities and designing aid programs. The problems rise both on the donor and on the recipient side.
There is only an indirect relationship between the people who provide the funds (taxpayers in donor countries) and the bene…ciaries of aid projects (the poors in recipient countries). In aid programs, there is a long and complex chain of principal-agent relationships, starting with the taxpayers that delegate authority to elected o¢ cials, who in turn become principals that delegate authority to a new set of agents, the chiefs of aid agencies, who delegate to agency employees and consultants.
In the recipient countries, there are similar relationships between citizens, their government, those who implement programs and interest groups that might be in favor or against the objectives of foreign assistance. The targets, the incentives and the information of the agents involved in the development strategies are not always in line with the objectives of the taxpayers or the bene…ciaries.
The principal-agent problem characterizes all public sector agencies and many private companies, but the international dimension and geographical distance between the original taxpayers and the ultimate recipients ampli…es the potential drawbacks. All aspects of aid delivery are a¤ected: program design, implementation, incentives, monitoring, evaluation, and allocation of funding. The instrument to overcome such limits is an adequate design of the institutions. The analysis of the motivations of aid donors becomes cru-cial for correctly dealing with the problem. In one approach, foreign aid is determined by the economic interests of powerful groups within donors. The idea is extended to multilateral aid by proxing for the quote of participation to multilateral agencies by the single governments. Another view explains aid, both bilateral and multilateral as an e¤ort to maximize bene…ts to donor states, deriving preferences for them from their situation in the international system. The …rst approach concentrates more on the pro…ts from trade and the economic gains from the access to natural resources, while the second fo-cuses on geopolitical features. An alternative perspective sees more generally aid as the outcome of bargaining among donor aid bureaucracies, multilateral
aid agencies and recipient government o¢ cials.
The Principal-Agent (P-A) problems do not emerge only on the donor-side. Recipient countries’ governments are often subject to the in‡uence of domestic special interest groups that act against the consensus of re-forms. Partly to overcome the P-A problem, donors often apply conditions on aid programs to push recipients to act in accordance with donors’interests. Donors’conditions on recipient policies have been subject to critiques. Policy conditionality is most often associated with the IMF and World Bank, but all donors use conditions to some extent. The rationale for economic policy conditions is well known: donors believe that certain policies are important determinant for growth and development and that only through their imple-mentation aid can have consistent positive returns. Aid has been used as a lever for policy reforms.
The problems with conditionality can be summed up in three aspects. First, it is di¢ cult sometimes to identify the most appropriate policy condi-tions to ensure sustained growth. Development theories suggested di¤erent approaches through time. In the 1950s and 1960s a state-led development strategy was dominant. In the 1970s it was replaced by basic human needs targets. The following decades the focus turned to a macroeconomic ap-proach centered on trade reforms and privatizations. From the mid-90s on, the importance of governance and institutional setting has become the core in accordance to the new …ndings of the growth literature. As a result, the list of conditions is constantly changing. The Washington Consensus menu of reforms greatly followed in the 1990s, for example, has been denied or modi…ed.
Second, imposing too many conditions or imposing none may lead to the same results. If a proper monitoring of the policy operational path is not undertaken or if the impositions of a large range of conditions only translates into the implementation of the less stringent ones the e¤ects of conditionality get lost.
Third, conditionality does not seem to work. Governments implement reforms when it is in their interests to do so. Domestic interest groups often in‡uence the government’s actions more powerfully than donors do. Many of them keep on disbursing aid even when recipients fail to meet conditions, sometimes in a repetitive fashion. What should be an incentive instrument for recipient is not so for donors. Their own internal incentives drive to keep on disbursing aid to support contractors and recipients that depend on it. Political and economic motivations mix with a “Samaritan’s dilemma”.
